CN108009647A - Equipment record processing method, device, computer equipment and storage medium - Google Patents
Equipment record processing method, device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN108009647A CN108009647A CN201711397566.2A CN201711397566A CN108009647A CN 108009647 A CN108009647 A CN 108009647A CN 201711397566 A CN201711397566 A CN 201711397566A CN 108009647 A CN108009647 A CN 108009647A
- Authority
- CN
- China
- Prior art keywords
- vector
- primary
- space
- map
- equipment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/20—Administration of product repair or maintenance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G07—CHECKING-DEVICES
- G07C—TIME OR ATTENDANCE REGISTERS; REGISTERING OR INDICATING THE WORKING OF MACHINES; GENERATING RANDOM NUMBERS; VOTING OR LOTTERY APPARATUS; ARRANGEMENTS, SYSTEMS OR APPARATUS FOR CHECKING NOT PROVIDED FOR ELSEWHERE
- G07C3/00—Registering or indicating the condition or the working of machines or other apparatus, other than vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Human Resources & Organizations (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- Strategic Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The present invention proposes a kind of equipment record processing method, device, computer equipment and storage medium, wherein, method includes：The first eigenvector in measurement value sensor generation primary vector space in being recorded according to each bar equipment, the maintenance in being recorded according to each bar equipment describe, the second feature vector in generation secondary vector space；According to the mapping relations between primary vector space and secondary vector space, the first map vector and the second map vector are determined；Choose the second map vector to be added in the primary vector collection comprising first eigenvector, choose the first map vector and be added in the secondary vector collection comprising second feature vector；Vector in primary vector collection and in secondary vector collection is clustered to obtain the first and second targets and is clustered；Cluster to first object and the second target cluster in similar cluster be combined, and determine equipment record belonging to classification.Pass through this method, by increasing capacitance it is possible to increase information content, makes up the drawbacks of available feature is insufficient in equipment record.
Description
Technical field
The present invention relates to technical field of information processing, more particularly to a kind of equipment record processing method, device, computer to set
Standby and storage medium.
Background technology
The same model device of same production firm production, may apply in each different place, with fire-fighting equipment or
Exemplified by person's air-conditioning equipment, almost each building can all lay fire-fighting equipment and air-conditioning equipment, the same all one's life in a urban district
The same model device of production factory production just has very much.These equipment need periodically or non-periodically to maintain in use, when
There are also needing to carry out maintenance it during failure, and after maintenance and repair by be filled in manually be used for this maintenance or maintain into
The maintenance description of row brief description., may for similar maintenance or maintenance but these maintenances are described due to being filled in manually
Different words is employed, how the maintenance description to these magnanimity is sorted out, to be adopted for of a sort maintenance or maintenance
It is described with unified form of presentation, in case it is technical problem urgently to be resolved hurrily subsequently to carry out accident analysis to use.
In the prior art, the method sorted out of maintenance description is mainly included method based on keyword and it is word-based to
Two kinds of the method for amount.Wherein, the method based on keyword is the keyword extracted in maintenance description, by with same keyword
Maintenance description merges；Method based on term vector is that the word during maintenance is described switchs to term vector, calculates the phase of term vector
Like degree, the corresponding maintenance description of the larger term vector of similarity is merged.
However, above-mentioned classifying method is realized based on text mining.Text mining not only needs to construct corpus, also needs
Chinese corpus is segmented, the accuracy sorted out using text mining mode to maintenance description is depended in the corpus
The abundant degree of maintenance description.In addition, participle also brings along more errors, and maintenance description mainly passes through manual entry, letter
Breath is more mixed and disorderly, and available feature is less.Directly maintenance description is being sorted out using text mining mode, and same class is being tieed up
When guarantor's description is described using unified presentation mode, describes to sort out inaccuracy there are maintenance, lead to not unified presentation mode
Technical problem.
The content of the invention
It is contemplated that solve at least some of the technical problems in related technologies.
For this reason, first purpose of the present invention is to propose a kind of equipment record processing method, to be remembered by collecting device
Corresponding device sensor data increase information content is recorded, makes up the drawbacks of available feature is insufficient in equipment record；By by equipment
Record data are converted to numerical value vector form, avoid text mining method from needing to construct the complex work of corpus；Pass through combination
Equipment record and sensing data is clustered, can caused by artificial subjective factor complex data effectively sorted out.
Second object of the present invention is to propose a kind of equipment record processing unit.
Third object of the present invention is to propose a kind of computer equipment.
Fourth object of the present invention is to propose a kind of non-transitorycomputer readable storage medium.
The 5th purpose of the present invention is to propose a kind of computer program product.
In order to achieve the above object, first aspect present invention embodiment proposes a kind of equipment record processing method, this method is used
In determining the classification belonging to each bar equipment record, described with the maintenance in the unified equipment record of the classification, including：
The first eigenvector in measurement value sensor generation primary vector space in being recorded according to each bar equipment, and root
Maintenance in being recorded according to each bar equipment describes, the second feature vector in generation secondary vector space；
According to the mapping relations between the primary vector space and the secondary vector space, determine each fisrt feature to
Measure the first map vector in secondary vector space, and second mapping of each second feature vector in primary vector space
Vector；
Choose second map vector to be added in the primary vector collection comprising the first eigenvector, and choose institute
The first map vector is stated to be added in the secondary vector collection comprising second feature vector；Choose the second map vector and its
The similarity vector that corresponding second feature vector is belonging respectively to different vector spaces clusters, and the first map vector of selection is right with it
The similarity vector that the first eigenvector answered is belonging respectively to different vector spaces clusters；
Vector in the primary vector collection is clustered to obtain first object and is clustered, and to the secondary vector collection
Interior vector, which is clustered to obtain the second target, to cluster；
The similar cluster to cluster to the first object in clustering with second target is combined, and according to each
The combination belonging to first eigenvector, second feature vector, the first map vector and the second map vector that equipment records, determines
Classification belonging to the equipment record.
The equipment of the embodiment of the present invention records processing method, is given birth to by the measurement value sensor in being recorded according to each bar equipment
First eigenvector into primary vector space, and the maintenance in each bar equipment record describes, generation secondary vector is empty
Interior second feature vector, according to the mapping relations between primary vector space and secondary vector space, determines that each first is special
First map vector of the sign vector in secondary vector space, and each second feature vector in primary vector space second
Map vector, chooses the second map vector and is added in the primary vector collection comprising first eigenvector, and chooses the first mapping
Vector is added in the secondary vector collection comprising second feature vector, and the vector in primary vector collection is clustered to obtain first
Target clusters, and the vector in secondary vector collection is clustered to obtain the second target clusters, and clusters to first object and the second mesh
Similar cluster during mark clusters is combined, and recorded according to each equipment first eigenvector, second feature vector, the
Combination belonging to one map vector and the second map vector, determines the classification belonging to equipment record, to realize to equipment record
Effectively sort out.Recording corresponding device sensor data by collecting device increases information content, compensate for use in equipment record
The drawbacks of feature is insufficient；By the way that equipment record data are converted to numerical value vector form, avoiding text mining method needs structure
Make the complex work of corpus；By bonding apparatus record and sensing data, same problem is clustered from two angles,
Can caused by artificial subjective factor complex data effectively sorted out, improve the accuracy rate of fault identification.
In order to achieve the above object, second aspect of the present invention embodiment proposes a kind of equipment record processing unit, for determining
Classification belonging to each bar equipment record, is described, which includes with the maintenance in the unified equipment record of the classification：
Generation module, for first in the measurement value sensor generation primary vector space in being recorded according to each bar equipment
Feature vector, and the maintenance in each bar equipment record describes, the second feature vector in generation secondary vector space；
Determining module, for according to the mapping relations between the primary vector space and the secondary vector space, really
Fixed first map vector of each first eigenvector in secondary vector space, and each second feature vector are empty in primary vector
The second interior map vector；
Module is chosen, is added to the primary vector for including the first eigenvector for choosing second map vector
In collection, and choose first map vector and be added in the secondary vector collection comprising second feature vector；The chosen
The similarity vector that the corresponding second feature vector of two map vectors is belonging respectively to different vector spaces clusters, and the first of selection
The similarity vector that the corresponding first eigenvector of map vector is belonging respectively to different vector spaces clusters；
Cluster module, clusters for being clustered to obtain first object to the vector in the primary vector collection, and right
Vector in the secondary vector collection, which is clustered to obtain the second target, to cluster；
Sort module, group is carried out for the similar cluster in clustering with second target that clusters to the first object
Close, and first eigenvector, second feature vector, the first map vector and the second map vector recorded according to each equipment
Affiliated combination, determines the classification belonging to the equipment record.
The equipment of the embodiment of the present invention record processing unit, is given birth to by the measurement value sensor in being recorded according to each bar equipment
First eigenvector into primary vector space, and the maintenance in each bar equipment record describes, generation secondary vector is empty
Interior second feature vector, according to the mapping relations between primary vector space and secondary vector space, determines that each first is special
First map vector of the sign vector in secondary vector space, and each second feature vector in primary vector space second
Map vector, chooses the second map vector and is added in the primary vector collection comprising first eigenvector, and chooses the first mapping
Vector is added in the secondary vector collection comprising second feature vector, and the vector in primary vector collection is clustered to obtain first
Target clusters, and the vector in secondary vector collection, which is clustered to obtain the second target, to cluster, and clusters to first object and the second target
Similar cluster in clustering is combined, and according to the first eigenvector of each equipment record, second feature vector, first
Combination belonging to map vector and the second map vector, determines the classification belonging to equipment record, has to realize to equipment record
Effect is sorted out.Recording corresponding device sensor data by collecting device increases information content, compensate for that spy can be used in equipment record
The drawbacks of sign deficiency；By the way that equipment record data are converted to numerical value vector form, avoiding text mining method needs to construct
The complex work of corpus；By bonding apparatus record and sensing data, same problem is clustered from two angles, energy
It is enough caused by artificial subjective factor complex data effectively sorted out, improve the accuracy rate of fault identification.
In order to achieve the above object, third aspect present invention embodiment proposes a kind of computer equipment, including：Memory, place
The computer program managed device and storage on a memory and can run on a processor, the processor perform the computer journey
During sequence, the equipment record processing method as described in first aspect embodiment is realized.
To achieve these goals, fourth aspect present invention embodiment proposes a kind of computer-readable storage of non-transitory
Medium, is stored thereon with computer program, and the equipment as described in first aspect embodiment is realized when which is executed by processor
Record processing method.
To achieve these goals, fifth aspect present invention embodiment proposes a kind of computer program product, when described
When instruction in computer program product is performed by processor, the equipment record processing side as described in first aspect embodiment is performed
Method.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obtain substantially, or recognized by the practice of the present invention.
Brief description of the drawings
Of the invention above-mentioned and/or additional aspect and advantage will become from the following description of the accompanying drawings of embodiments
Substantially and it is readily appreciated that, wherein：
The flow diagram for the equipment record processing method that Fig. 1 is provided by the embodiment of the present invention one；
Fig. 2 is the process schematic that feature vector is converted to map vector；
The flow diagram for the equipment record processing method that Fig. 3 is provided by the embodiment of the present invention two；
The flow diagram for the equipment record processing method that Fig. 4 is provided by the embodiment of the present invention three；
The flow diagram for the equipment record processing method that Fig. 5 is provided by the embodiment of the present invention four；
The structure diagram for the equipment record processing unit that Fig. 6 is provided by the embodiment of the present invention one；
The structure diagram for the equipment record processing unit that Fig. 7 is provided by the embodiment of the present invention two；
The structure diagram for the equipment record processing unit that Fig. 8 is provided by the embodiment of the present invention three；
The structure diagram for the equipment record processing unit that Fig. 9 is provided by the embodiment of the present invention four；And
Figure 10 is the structure diagram for the computer equipment that one embodiment of the invention proposes.
Embodiment
The embodiment of the present invention is described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or has the function of same or like element.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and is not considered as limiting the invention.
Below with reference to the accompanying drawings the equipment of the embodiment of the present invention record processing method, device, computer equipment and storage are described
Medium.
The flow diagram for the equipment record processing method that Fig. 1 is provided by the embodiment of the present invention one, this method can be true
Classification belonging to fixed each bar equipment record.Since each equipment record includes two parts data, a part is used for instruction equipment
In the measurement value sensor that collects in fixation duration of each sensor before failure occurs or before maintenance, another part is artificial
The maintenance description filled in, language description is carried out for the maintenance of the failure to generation or progress.Therefore, in definite each bar equipment
After classification belonging to record, it can be carried out to belonging to the maintenance description in of a sort equipment record using unified form of presentation
Statement, unifies maintenance description, in case subsequently carrying out the purpose of accident analysis so as to reach according to classification.
As shown in Figure 1, equipment record processing method comprises the following steps：
Step 101, the fisrt feature in measurement value sensor generation primary vector space in being recorded according to each bar equipment
Vector, and the maintenance in each bar equipment record describes, the second feature vector in generation secondary vector space.
In the present embodiment, the corresponding primary fault maintenance of each equipment record or primary equipment maintenance, equipment record
Specifically include measurement value sensor and the maintenance description of each sensor.
Wherein, measurement value sensor is collected in the fixation duration before breaking down or carrying out corrective maintenance
, such as：Equipment includes the sensor for being respectively used to multiple parameters such as time of measuring, temperature, pressure, rotating speed, voltage, electric current,
Wherein, the measurement value sensor of time-parameters is 2017-12-13 10:30:45；The measurement value sensor of temperature parameters is 60；
The measurement value sensor of pressure variable is 40；The measurement value sensor of rotating speed parameter is 100；The measurement value sensor of voltage parameter
For 220；The measurement value sensor of current parameter is 40.
Multiple measurement value sensors in being recorded to each equipment, vectorization is carried out in primary vector space, is obtained
The first eigenvector of this equipment record.For example, foregoing multiple measurement value sensors can in primary vector space into
Row vector, obtains the first eigenvector (2017-12-13 10 of 1*6 matrix forms:30:45,60,40,100,220,40).
Primary vector space defines the implication (time, temperature, pressure, rotating speed, voltage, electric current) of each element.It should be noted that
Multiple measurement value sensors of each bar equipment record, vectorization is carried out in same primary vector space, obtains corresponding the
One feature vector.
Maintenance in being recorded to each equipment describes, main by being used as in being described to maintenance in secondary vector space
The notional word of feature carries out vectorization, obtains the second feature vector in secondary vector space, and second feature vector here is specific
It can be the digitized vector obtained to maintenance description after dummy argument, word2vector processing.
Example is turned to dummy argument, count in whole maintenances records as " fault type ", " failure rank ", " settling mode ",
The notional word of the value of " failure cause " this four features, it is assumed that the value of these features is at least one in 10 possible notional words
It is a, then the matrix form that the second feature vector of maintenance record is 4*10, secondary vector space define the row instruction of matrix
Feature, and the notional word indicated by matrix column.In each maintenance records corresponding matrix, each element uses 0 or 1 table
Show, element is in 1 representing matrix corresponding maintenance record, which is expert at the feature of instruction, and value is the element column
The notional word of instruction；Element is in zero representing matrix corresponding maintenance record, which is expert at the feature of instruction, and value is not this
The notional word of element column instruction.
Assuming that maintenance is recorded as " air-conditioning does not freeze, and part aging causes, normal operation after renewal part ".Fault type is
Do not freeze, failure rank is to need to repair, and settling mode is renewal part, and failure cause is part aging.Maintenance record corresponds to
Matrix in, following element value be 1, remaining element is zero：
Feature " fault type " correspond to row, notional word " do not freeze " respective column element value be 1；
Feature " failure rank " corresponds to row, and the element value of notional word " maintenance " respective column is 1；
Feature " settling mode " corresponds to row, and the element value of notional word " renewal part " respective column is 1；
Feature " failure cause " corresponds to row, and the element value of notional word " part aging " respective column is 1.
Step 102, according to the mapping relations between primary vector space and secondary vector space, determine each fisrt feature to
Measure the first map vector in secondary vector space, and second mapping of each second feature vector in primary vector space
Vector.
What is stored in primary vector space is the first eigenvector of the measurement value sensor generation in being recorded according to equipment,
What is stored in secondary vector space is that the second feature that the maintenance description in being recorded according to equipment generates is vectorial, primary vector space
There are mapping relations between secondary vector space, both can mutually be changed.
Here conversion is to carry out vector transformation realization by formula.Specifically, what is used is in linear algebra
Matrix multiplication principle.Such as：The matrix A of one 2*3 is multiplied with the matrix V of a 3*4, obtains the matrix V of a 2*4 ', 3* at this time
Matrix V in 4 vector spaces is converted into the matrix V in 2*4 vector spaces by matrix A '.Here matrix A be exactly from 3*4 to
The V of quantity space is mapped to the mapping relations of the V ' of 2*4 vector spaces., can be according to primary vector space and in the present embodiment
Mapping relations between two vector spaces, for each first eigenvector, determine first eigenvector in secondary vector space
The first interior map vector, and, for each second feature vector, determine second feature vector in primary vector space
Second map vector.
Specifically, in order to enable the difference vector in primary vector space is still difference after secondary vector space is mapped to
Vector, similarly, the difference vector in secondary vector space is still different vectors after primary vector space is mapped to, and mapping is closed
System is frequently not the relation of foregoing relatively simple V '=AV, and is set to the following formula (1) form, that is to say, that true
When determining the map vector of feature vector, it can be calculated according to equation below (1).
V'=σ (w*V+b) (1)
Wherein, w is coefficient, and b is biasing, and V represents feature vector, and V ' represents map vector, and σ can be sigmoid functions.
In formula (1), the value of w and b can be by determining, when being iterated calculating after successive ignition, can be with friendship
Entropy formula is pitched as loss function, shown in loss function such as formula (2), and it is possible to set threshold value for loss function or set
Maximum iteration is put, using the condition as iteration ends.
Wherein, xjRepresent j-th of first eigenvector,Represent xjEstimate, be in an iterative process by xjAs V
Value substitute into formula (1) value of obtained V ', n represents the number of first eigenvector in primary vector space；yjTable
Show j-th of second feature vector,Represent yjEstimate, be in an iterative process by xjValue as V substitutes into formula (1)
In the obtained value of V ', m represents the number of second feature vector in secondary vector space.Loss_s2d represent fisrt feature to
Measure the loss function changed to the first map vector；Loss_d2s represents the damage that second feature vector is changed to the second map vector
Lose function.
It is iterated to calculate using the loss function shown in formula (2) and determines shown in the formula such as formula (3) of parameter w and b.
Wherein,W and b in formula (1) when first eigenvector is converted to the first map vector during expression current iteration
Value, θsW and b in formula (1) when first eigenvector is converted to the first map vector in the last iterative process of expression
Value；Represent the value of second feature vector w and b in formula (1) when being converted to the second map vector during current iteration,
θdRepresent the value of w and b in formula (1) when second feature vector is converted to the second map vector in last iterative process.lr
Represent learning rate (learning rate), determine that parameter is moved to the speed speed of optimal value, lr is excessive may to cause parameter
Cross optimal value, and lr is too small that algorithm may be caused can not to restrain for a long time.Lr value ranges are (0,1), in practical applications,
Usually take 10-4~10-3。
Fig. 2 is the process schematic that feature vector is converted to map vector.As shown in Fig. 2, repeatedly change by foregoing
After process, it may be determined that go out first eigenvector to the first map vector convert when w and b value, and second feature
Vector to the second map vector convert when w and b value, the value of w and b in the case of both different conversions is difference
's.By first eigenvector to the first map vector convert when w and b value substitute into formula (1) obtain primary vector space
To the mapping relations of secondary vector spatial transformation, by second feature vector to the second map vector convert when w and b value generation
Enter formula (1) and obtain secondary vector space to the mapping relations of primary vector spatial transformation.
According to both the above mapping relations, (DS (v can be denoted as to primary vector spaces)) in each fisrt feature to
Amount (is denoted as vs) changed, obtain the first map vector and (be denoted as vd'), and (DS (v are denoted as to secondary vector spaced)) in
Second feature vector (be denoted as vd) changed, obtain the second map vector and (be denoted as vs’)。
Step 103, choose the second map vector to be added in the primary vector collection comprising first eigenvector, and choose the
One map vector is added in the secondary vector collection comprising second feature vector.
Wherein, the corresponding second feature vector of the second map vector of selection is belonging respectively to the phase of different vector spaces
Cluster like vector, the first eigenvector that the first map vector of selection is corresponding is belonging respectively to the similar of different vector spaces
Vector clusters.
In the present embodiment, relevant clustering algorithm (such as k-means algorithms) can be used in primary vector space
Whole first eigenvectors are clustered, and whole second feature vectors in secondary vector space are clustered, and obtain two
Group cluster is as a result, clustering of being obtained in primary vector space is labeled asCluster what is obtained in secondary vector space
It is labeled asWherein, it is sameThe first eigenvector included in clustering is to be in similar state according to equipment
When measurement value sensor generation, it is sameThe second feature vector included in clustering is to be in similar according to equipment
What maintenance record during state generated.That is, same cluster has corresponded to similar equipment state, difference, which clusters, have been corresponded to not
Same equipment state.
Equipment can produce maintenance record when carrying out maintenance maintenance, while also have measurement value sensor, thus have
Two kinds of data types.In order to obtain multiple equipment generation measurement value sensor and with these measurement value sensors at least
The maintenance record of one corresponding like device state, it is necessary to obtain the set for recording measurement value sensor first, and then, if
Maintenance record equipment state corresponding with least one measurement value sensor in this set is similar, then records this maintenance
It is added in this set.Specifically, the second map vector is calculated to existIn belonging to cluster, and calculate and first reflect
Directive amount existsIn belonging to cluster.If the second map vector existsIn belonging to cluster and corresponding second is special
Sign vector existsIn belonging to cluster be that similarity vector clusters, then by the second map vector be added to comprising fisrt feature to
In the primary vector collection of amount.Similarly, in order to obtain multiple equipment generation maintenance record and with these maintenances record in extremely
The measurement value sensor of a few corresponding like device state, obtains the set for recording maintenance record first, and then, if passing
It is similar that sensor measured value with least one maintenance in this set records corresponding equipment state, then measures this sensor
Value is added in this set.Specifically, if the first map vector existsIn belonging to cluster and corresponding first
Feature vector existsIn belonging to cluster be that similarity vector clusters, then the first map vector is added to comprising second feature
In the secondary vector collection of vector.
It is added to herein it should be noted that choosing the second map vector in primary vector collection, and chooses the first mapping
Vector will be provided added to the specific implementation process in secondary vector collection in subsequent content, to avoid repeating, not made herein in detail
Thin description.
Step 104, the vector in primary vector collection is clustered to obtain first object and is clustered, and to secondary vector collection
Interior vector, which is clustered to obtain the second target, to cluster.
In previous step, primary vector collection and secondary vector collection are generated, specifically, primary vector collection is based on first
Vector space, for indicate multiple equipment produce measurement value sensor and with it is at least one in these measurement value sensors
The maintenance record of corresponding like device state；Secondary vector collection is based on secondary vector space, for indicating multiple equipment generation
Maintenance records and the measurement value sensor with least one corresponding like device state in these maintenances record.Can be again
Primary vector collection and secondary vector collection are clustered respectively using relevant clustering method, first object is obtained and clusters and second
Target clusters.Here, same first object clusters maps comprising first eigenvector similar in primary vector space and second
Vector, the measurement value sensor of first eigenvector instruction, corresponding similar sets with the maintenance record of the second map vector instruction
Standby state；Same second target clusters comprising the first map vector of second feature vector sum similar in secondary vector space, the
The measurement value sensor of one map vector instruction, similar equipment state corresponding with the maintenance record of the second map vector instruction.
Step 105, cluster to first object and the second target cluster in similar cluster be combined, and according to each
The combination belonging to first eigenvector, second feature vector, the first map vector and the second map vector that equipment records, determines
Classification belonging to equipment record.
Specifically, obtaining first object and clustering (to be denoted as) and the second target cluster and (be denoted as) after,
The similarity distance between each barycenter to cluster during two targets cluster can be first calculated, different vector spaces is specifically calculated and clusters
Barycenter between similarity distance can be in the following ways：
After the centroid vector that first object is clustered maps to secondary vector space, it can calculate and in second space second
The first distance that target clusters between centroid vector；And map to primary vector in the centroid vector that the second target clusters
Behind space, calculate and primary vector space in the second distance that clusters between centroid vector of first object.First distance is added
Second distance, obtains the similarity distance between the barycenter that different vector spaces cluster.Shown in calculation formula such as formula (4).
Sim (x, y)=sim (x → y)+sim (y → x) (4)
Wherein,Represent the barycenter that the second target clusters,Table
Show the barycenter that first object clusters.After sim (y → x) represents that the centroid vector that first object clusters maps to secondary vector space,
The first distance between the centroid vector that clusters with the second target in second space；Sim (x → y) represents the matter that the second target clusters
After Heart vector maps to primary vector space, between the centroid vector that clusters with first object in primary vector space second away from
From；Similarity distance between the barycenter that the first object of sim (x, y) difference vector spaces clusters and the second target clusters.
, can be with preset threshold value in the present embodiment, and the similarity distance that two are clustered is less than two of threshold value and clusters
It is combined.Assuming that first object clusters after the combination of the similar cluster in clustering with the second target, obtain z and cluster, then it is right
Z classification is answered, is denoted as c ∈ (c1,c2,…cz).Different classes of to have corresponded to different equipment states, same category has corresponded to similar
Equipment state.
Each first eigenvector, second feature vector, the first map vector and the second map vector are belonging respectively to class
Other ciIn one kind, corresponding classification is recorded to the equipment and is counted, selects the most classification of occurrence number as the equipment
Classification belonging to record.If the number that each classification occurs is identical, the classification of confidence level maximum is determined as equipment note
Classification belonging to record.
Such as：To an equipment record (equipment is in a certain equipment state at this time), including maintenance record, while also wrap
Include measurement value sensor during equipment operation.Maintenance record can obtain the of secondary vector space by vectorization procedure above
Two feature vectors and second map vector in primary vector space；Measurement value sensor can obtain by vectorization procedure above
The first eigenvector in primary vector space and first map vector in secondary vector space.Foregoing 4 vectors are obtained, to the greatest extent
Pipe is in different vector spaces, but is all various forms of descriptions to same equipment state, by gathering in different vector spaces
After class, it is combined in the similar cluster to different vector spaces, can obtains corresponding to same category of combination, according to each vectorial institute
The classification of category come judge the equipment record belonging to classification.
After the classification of each equipment record is determined, the equipment record in the category is produced under like device state
, the highest maintenance description of the frequency of occurrences, is named the category, using the category in being recorded according to same category of equipment
Naming method name same category in each bar equipment record maintenance description.
The equipment record processing method of the present embodiment, passes through the measurement value sensor generation in being recorded according to each article of equipment the
First eigenvector in one vector space, and the maintenance in each bar equipment record describes, in generation secondary vector space
Second feature vector, according to the mapping relations between primary vector space and secondary vector space, determine each fisrt feature to
Measure the first map vector in secondary vector space, and second mapping of each second feature vector in primary vector space
Vector, chooses the second map vector and is added in the primary vector collection comprising first eigenvector, and chooses the first map vector
It is added in the secondary vector collection comprising second feature vector, the vector in primary vector collection is clustered to obtain first object
Cluster, the vector in secondary vector collection is clustered to obtain the second target cluster, cluster to first object and the second target is gathered
Similar cluster in cluster is combined, and is reflected according to the first eigenvector of each equipment record, second feature vector, first
The combination belonging to the second map vector of vector sum is penetrated, determines the classification belonging to equipment record, to realize to the effective of equipment record
Sort out.Recording corresponding device sensor data by collecting device increases information content, compensate for available feature in equipment record
The drawbacks of insufficient；By the way that equipment record data are converted to numerical value vector form, avoiding text mining method needs to construct language
Expect the complex work in storehouse；By bonding apparatus record and sensing data, same problem is clustered from two angles, can
Caused by artificial subjective factor complex data effectively sorted out, improve the accuracy rate of fault identification.
In order to clearly describe the measurement value sensor generation in being recorded in previous embodiment according to each article of equipment the
First eigenvector in one vector space, and the maintenance in each bar equipment record describes, in generation secondary vector space
Second feature vector specific implementation process, the embodiment of the present invention proposes another equipment record processing method, and Fig. 3 is this
The flow diagram for the equipment record processing method that inventive embodiments two are provided.
As shown in figure 3, on the basis of embodiment as shown in Figure 1, step 101 may comprise steps of：
Step 201, a dimension using each measurement value sensor as vector, carries out vectorization, and it is special to obtain first
Sign vector.
Wherein, measurement value sensor can for example include at least one in temperature value, rotating speed and pressure value.
In the present embodiment, it is T time point that can preset time span, is recorded for each equipment, utilizes this
T time point corresponding measurement value sensor before the fault time point of equipment record, forms the matrix M=(x of T × St,s),
t<T,s<S, wherein, the row of matrix M is used for instruction time point, and row are used for the corresponding sensor of indication sensor measured value, and S is should
The number of probes of equipment, that is, the dimension of the measurement value sensor collected, a kind of measured value correspond to a dimension.
For the matrix M of measurement value sensor composition, can be averaged by the way of averaging to each column element,
Value using the average value of each row as the element of respective column in first eigenvector, is calculated first eigenvector and (uses vs
Represent), shown in calculation formula such as formula (5).
Wherein, biThe measured value at the T time point of sensor of the i-th row is averaged in (i=1,2 ..., S) representing matrix M
Value.
In a kind of possible implementation of the embodiment of the present invention, before vectorization is carried out to measurement value sensor, also
It can first be averaged to same sensor in measured value at different moments, obtain a kind of measurement value sensor, and then using
The measurement value sensor of each dimension of gained, generates first eigenvector after value calculates.
Step 202, using each maintenance describe as vector a dimension, progress vectorization, obtain second feature to
Amount.
Wherein, the maintenance description in maintenance record can for example include maintenance time, device type, fault category, failure
It is at least one in rank, vendor name and the source of failure.
In the present embodiment, the maintenance in being recorded for each equipment describes, and carries out vectorization using dummy argument mode, obtains
To second feature vector, v is denoted asd=[a1,a2,...,an], wherein, n is vector dimension, represents that maintenance is retouched in equipment record
The species number stated.Specific dummy argument mode repeats no more this in the present embodiment referring to previous embodiment.
The equipment record processing method of the present embodiment, corresponding measurement value sensor, Neng Gouzeng are recorded by collecting device
Add information content, make up the drawbacks of equipment recording feature is insufficient；By the way that maintenance description is converted to feature vector, text digging is avoided
Pick needs to build the complex work of corpus, reduces workload.
It should be noted that abovementioned steps 201 and step 202 can be performed successively, or parallel, the embodiment of the present invention pair
It is not construed as limiting in the execution sequence of step 201 and step 202.In Fig. 3 step 201 and step are shown with different connection modes
202 execution sequence, wherein, the arrow on the right is directed toward step 202 by step 201 and represents to perform step again after first carrying out step 201
Rapid 202；The arrow on the left side is directed toward step 201 by step 202 and represents to perform step 201 again after first carrying out step 202；Middle is double
Represent that step 201 and step 202 perform parallel to arrow.
It is added to clearly describe to choose the second map vector in previous embodiment comprising first eigenvector
Primary vector collection in, and choose the first map vector and be added to specific reality in the secondary vector collection comprising second feature vector
Existing process, the embodiment of the present invention propose another equipment record processing method, and Fig. 4 is set by what the embodiment of the present invention three provided
The flow diagram of note processing method.
As shown in figure 4, have been described above the basis of each step and the physical meaning of formula in embodiment as shown in Figure 1
On, emphasis is subjected to careful description to calculating process in following embodiment.Step 103 may comprise steps of：
Step 301, in primary vector space, first eigenvector is clustered to obtain each primary vector clustered；
In secondary vector space, second feature vector is clustered to obtain each secondary vector clustered.
In the present embodiment, first eigenvector and second feature vector can be carried out respectively using relevant clustering algorithm
Cluster, such as, k-means algorithms can be used to be clustered to obtain each primary vector to first eigenvector and cluster, and it is right
Second feature vector, which is clustered to obtain each secondary vector, to cluster.
Step 302, according to the mapping relations between primary vector space and secondary vector space, determine that primary vector clusters
And secondary vector cluster between similitude.
Specifically, according to the mapping relations between primary vector space and secondary vector space, determine that primary vector clusters
And secondary vector cluster between similitude when, can first obtain mapping between primary vector space and secondary vector space and close
System.
In the present embodiment, when obtaining the mapping relations between primary vector space and secondary vector space, it can obtain respectively
Primary vector space is taken to the mapping relations of secondary vector space reflection, and secondary vector space is to primary vector space reflection
Mapping relations.
Specifically, can be by first eigenvector vsSubstitute into transfer function V'=σ (w*V+b) and be iterated computing, with
The first map vector v when taking different value to w and bd', and according to the first map vector vd' and corresponding second feature vector
vdDetermine loss function value；Wherein, σ is S type functions, such as, σ can be sigmoid functions.When loss function value is less than
Threshold value or while reaching maximum iteration, stop iteration, by w and b values substitution transfer function when stopping iteration, as first
Vector space is converted into the mapping relations in secondary vector space.
Obtain mapping relations from secondary vector space to primary vector space reflection when, can be by second feature vector vdGeneration
Enter transfer function V'=σ (w*V+b) and be iterated computing, to obtain the second map vector v when w and b takes different values', and
According to the second map vector vs' and corresponding first eigenvector vsDetermine loss function value；Wherein, σ is S type functions.Work as damage
Lose function value to be less than threshold value or stop iteration when reaching maximum iteration, w and b values when stopping iteration being substituted into conversion letter
Number, as the mapping relations that secondary vector spatial transformation is primary vector space.
, can basis after obtaining the mapping relations between primary vector space and secondary vector space in the present embodiment
Mapping relations, the centroid vector that primary vector is clustered map to secondary vector space, and the barycenter that secondary vector is clustered
For DUAL PROBLEMS OF VECTOR MAPPING to primary vector space, mapping process may refer to the associated description of previous embodiment, and as shown in Figure 2 reflect
Penetrate process schematic.
Further, after centroid vector primary vector to be clustered maps to secondary vector space, can calculate and second
Secondary vector clusters the first distance between centroid vector in vector space；And in the centroid vector that secondary vector clusters
After mapping to primary vector space, calculate and primary vector space in the second distance that clusters between centroid vector of primary vector.
Wherein it is possible to the first distance and second distance are calculated according to Euclidean distance formula.
Finally, according to the first distance and second distance, it may be determined that between primary vector clusters and secondary vector clusters
Similitude.For example two closer to the distance in the range of distance threshold can be clustered and be determined as existing with preset distance threshold value
Similitude.
Step 303, if the primary vector belonging to first eigenvector clusters, the first mapping corresponding with first eigenvector
Secondary vector belonging to vector cluster between there are similitude, the first map vector is added in secondary vector collection.
Step 304, if the secondary vector belonging to second feature vector clusters, the second mapping corresponding with second feature vector
Primary vector belonging to vector cluster between there are similitude, the second map vector is added in primary vector collection.
In the present embodiment, for each first eigenvector and corresponding first map vector, if fisrt feature
Primary vector belonging to vector cluster with the secondary vector belonging to the first map vector cluster between there are similitude, then by first
Image vector is added in secondary vector collection.Similarly, for each second feature it is vectorial and corresponding second map to
Amount, if the secondary vector belonging to second feature vector cluster with the primary vector belonging to the second map vector cluster between exist
Second image vector, then be added in primary vector collection by similitude.
The equipment record processing method of the present embodiment, by gathering respectively to first eigenvector and second feature vector
Class obtains primary vector and clusters to cluster with secondary vector, is closed according to the mapping between primary vector space and secondary vector space
System, determine primary vector cluster and secondary vector cluster between similitude, and in the primary vector belonging to first eigenvector
The secondary vector belonging to the first map vector corresponding with first eigenvector that clusters cluster between there are during similitude, by first
Map vector is added in secondary vector collection, clusters in the secondary vector belonging to second feature vector corresponding with second feature vector
The second map vector belonging to primary vector cluster between there are similitude, the second map vector is added to primary vector collection
It is interior, by increasing capacitance it is possible to increase information content, increases characteristic information, lays the foundation for the accurate fault identification that carries out.
In a kind of possible implementation of the embodiment of the present invention, as shown in figure 5, on the basis of embodiment as shown in Figure 1
On, before step 105, further include：
Step 401, the mapping relations between primary vector space and secondary vector space are obtained.
Step 402, according to mapping relations, after the centroid vector that first object is clustered maps to secondary vector space, meter
Calculate and secondary vector space in the 3rd distance that clusters between centroid vector of the second target.
Step 403, according to mapping relations, after the centroid vector that the second target clusters is mapped to primary vector space, meter
Calculate and primary vector space in the 4th distance that clusters between centroid vector of first object.
It should be noted that the description to step 401- steps 403, may refer in previous embodiment to step 302
Description, its realization principle is similar, and details are not described herein again.
Step 404, according to the 3rd distance and the 4th distance, determine first object cluster and the second target cluster between phase
Like property.
In the present embodiment, it is calculated after the first distance and second distance, the first distance and second distance can be calculated
The sum of distance, by the sum of gained distance compared with default threshold value, and when the sum of distance is more than threshold value, determine first
There are similitude between target clusters and the second target clusters.
The equipment record processing method of the present embodiment, by obtaining reflecting between primary vector space and secondary vector space
Penetrate relation, after the centroid vector that first object is clustered according to mapping relations maps to secondary vector space, calculate with second to
The second target clusters the 3rd distance between centroid vector in quantity space, and the centroid vector that the second target is clustered maps to
Behind primary vector space, calculate and primary vector space in the 4th distance that clusters between centroid vector of first object, according to the
Three distances and the 4th distance determine first object cluster with the second target cluster between similitude, with according to similitude to first
Similar cluster between target clusters and the second target clusters is combined, by increasing capacitance it is possible to increase cluster number, and then increases classification
Number, to identify that further types of failure lays the foundation.
In order to realize above-described embodiment, the present invention also proposes a kind of equipment record processing unit.
The structure diagram for the equipment record processing unit that Fig. 6 is provided by the embodiment of the present invention one, the device are used for true
Classification belonging to fixed each bar equipment record, is described with the maintenance in being recorded according to classification Unified Device.
As shown in fig. 6, the equipment record processing unit 60 includes：Generation module 610, determining module 620, choose module
630th, cluster module 640, and sort module 650.Wherein,
Generation module 610, in the measurement value sensor generation primary vector space in being recorded according to each bar equipment
First eigenvector, and the maintenance in each bar equipment record describes, the second feature vector in generation secondary vector space.
Determining module 620, for according to the mapping relations between primary vector space and secondary vector space, determining each
First map vector of one feature vector in secondary vector space, and each second feature vector is in primary vector space
Second map vector.
Module 630 is chosen, is added to for choosing the second map vector in the primary vector collection comprising first eigenvector,
And choose the first map vector and be added in the secondary vector collection comprising second feature vector；Wherein, the second of selection map to
Measure corresponding second feature vector and be belonging respectively to the similarity vectors of different vector spaces and cluster, the first map vector of selection
The similarity vector that corresponding first eigenvector is belonging respectively to different vector spaces clusters.
Cluster module 640, clusters for being clustered to obtain first object to the vector in primary vector collection, and to
Vector in two vector sets, which is clustered to obtain the second target, to cluster.
Sort module 650, for cluster to first object and the second target cluster in similar cluster be combined, and root
According to belonging to the first eigenvector of each equipment record, second feature vector, the first map vector and the second map vector
Combination, determines the classification belonging to equipment record.
Further, in a kind of possible implementation of the embodiment of the present invention, as shown in fig. 7, implementing as shown in Figure 6
On the basis of example, generation module 610 includes：
First generation unit 611, for a dimension using each measurement value sensor as vector, into row vector
Change, obtain first eigenvector；Measurement value sensor includes at least one in temperature value, rotating speed and pressure value.
Alternatively, in a kind of possible implementation of the embodiment of the present invention, the first generation unit 611 is special in generation first
It before sign vector, can also first be averaged to same sensor in measured value at different moments, obtain a kind of sensor measurement
Value.
Second generation unit 612, for a dimension using each maintenance description as vector, carries out vectorization, obtains
To second feature vector；Maintenance description includes maintenance time, device type, fault category, failure rank, vendor name and failure
It is at least one in source.
In a kind of possible implementation of the embodiment of the present invention, as shown in figure 8, on the basis of embodiment as shown in Figure 6
On, choosing module 630 includes：
Cluster cell 631, in primary vector space, being clustered to obtain each primary vector to first eigenvector
Cluster；In secondary vector space, second feature vector is clustered to obtain each secondary vector clustered.
Determination unit 632, for according to the mapping relations between primary vector space and secondary vector space, determining first
Similitude between vector clusters and secondary vector clusters.
Specifically, it is determined that unit 632 can first obtain the mapping relations between primary vector space and secondary vector space.
, can be by first eigenvector when determination unit 632 obtains the mapping relations between primary vector space and secondary vector space
vsSubstitute into transfer function V'=σ (w*V+b) and be iterated computing, to obtain the first map vector v when w and b takes different valued',
And according to the first map vector vd' and corresponding second feature vector vdDetermine loss function value；σ is S type functions；Work as damage
Lose function value to be less than threshold value or stop iteration when reaching maximum iteration, w and b values when stopping iteration being substituted into conversion letter
Number, as the mapping relations that primary vector spatial transformation is secondary vector space；And by second feature vector vdSubstitute into conversion
Function V'=σ (w*V+b) are iterated computing, to obtain the second map vector v when w and b takes different values', and according to
Two map vector vs' and corresponding first eigenvector vsDetermine loss function value；σ is S type functions；When loss function value
Stop iteration less than threshold value or when reaching maximum iteration, w and b values substitution transfer function when will stop iteration, is used as the
Two vector spaces are converted into the mapping relations in primary vector space.
Further, it is determined that unit 632 obtain mapping relations between primary vector space and secondary vector space it
Afterwards, can according to mapping relations, after the centroid vector that primary vector is clustered maps to secondary vector space, calculate with second to
Secondary vector clusters the first distance between centroid vector in quantity space；And according to mapping relations, secondary vector is clustered
After centroid vector maps to primary vector space, calculate and primary vector space in primary vector cluster between centroid vector the
Two distances.Finally, determination unit 632 determines that primary vector clusters and clusters with secondary vector according to the first distance and second distance
Between similitude.
Adding device 633, it is corresponding with first eigenvector for clustering in the primary vector belonging to first eigenvector
Secondary vector belonging to first map vector cluster between there are during similitude, the first map vector is added to secondary vector collection
It is interior；And cluster in the secondary vector belonging to second feature vector, belonging to the second map vector corresponding with second feature vector
Primary vector cluster between there are during similitude, the second map vector is added in primary vector collection.
In a kind of possible implementation of the embodiment of the present invention, as shown in figure 9, in the base of embodiment as shown in Figure 6
On plinth, which can also include：
Processing module 660, for cluster to first object and the second target cluster in similar cluster be combined before,
Obtain the mapping relations between primary vector space and secondary vector space；According to mapping relations, the matter that first object is clustered
After Heart vector maps to secondary vector space, calculate and secondary vector space in the second target cluster between centroid vector the 3rd
Distance；According to mapping relations, after the centroid vector that the second target clusters is mapped to primary vector space, calculating and primary vector
First object clusters the 4th distance between centroid vector in space；According to the 3rd distance and the 4th distance, first object is determined
Cluster and the second target cluster between similitude.
It should be noted that the foregoing explanation to equipment record processing method embodiment is also applied for the embodiment
Equipment record processing unit, its realization principle is similar, and details are not described herein again.
The equipment record processing unit of the present embodiment, passes through the measurement value sensor generation in being recorded according to each article of equipment the
First eigenvector in one vector space, and the maintenance in each bar equipment record describes, in generation secondary vector space
Second feature vector, according to the mapping relations between primary vector space and secondary vector space, determine each fisrt feature to
Measure the first map vector in secondary vector space, and second mapping of each second feature vector in primary vector space
Vector, chooses the second map vector and is added in the primary vector collection comprising first eigenvector, and chooses the first map vector
It is added in the secondary vector collection comprising second feature vector, the vector in primary vector collection is clustered to obtain first object
Cluster, the vector in secondary vector collection, which is clustered to obtain the second target, to cluster, and clusters to first object and the second target clusters
In similar cluster be combined, and according to each equipment record first eigenvector, second feature vector, first mapping
Combination belonging to the second map vector of vector sum, determines the classification belonging to equipment record, to realize effectively returning to equipment record
Class.Recording corresponding device sensor data by collecting device increases information content, compensate in equipment record available feature not
The drawbacks of sufficient；By the way that equipment record data are converted to numerical value vector form, avoiding text mining method needs to construct language material
The complex work in storehouse；By bonding apparatus record and sensing data, same problem is clustered from two angles, can be right
Complex data is effectively sorted out caused by artificial subjective factor, improves the accuracy rate of fault identification.
In order to realize above-described embodiment, the present invention also proposes a kind of computer equipment.
Figure 10 is the structure diagram for the computer equipment that one embodiment of the invention proposes.As shown in Figure 10, the computer
Equipment 100 includes：Memory 110, processor 120 and it is stored in the calculating that can be run on memory 110 and on processor 120
Machine program 130, when processor 120 performs computer program 130, realizes equipment record processing side as in the foregoing embodiment
Method.
In order to realize above-described embodiment, the present invention also proposes a kind of non-transitorycomputer readable storage medium, deposits thereon
Computer program is contained, equipment record processing method as in the foregoing embodiment is realized when which is executed by processor.
In order to realize above-described embodiment, the present invention also proposes a kind of computer program product, when in computer program product
Instruction when being performed by processor, perform equipment record processing method as in the foregoing embodiment.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or the spy for combining the embodiment or example description
Point is contained at least one embodiment of the present invention or example.In the present specification, schematic expression of the above terms is not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
Combined in an appropriate manner in one or more embodiments or example.In addition, without conflicting with each other, the skill of this area
Art personnel can be tied the different embodiments or example described in this specification and different embodiments or exemplary feature
Close and combine.
In addition, term " first ", " second " are only used for description purpose, and it is not intended that instruction or hint relative importance
Or the implicit quantity for indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, " multiple " are meant that at least two, such as two, three
It is a etc., unless otherwise specifically defined.
Any process or method described otherwise above description in flow chart or herein is construed as, and represents to include
Module, fragment or the portion of the code of the executable instruction of one or more the step of being used for realization custom logic function or process
Point, and the scope of the preferred embodiment of the present invention includes other realization, wherein can not press shown or discuss suitable
Sequence, including according to involved function by it is basic at the same time in the way of or in the opposite order, carry out perform function, this should be of the invention
Embodiment person of ordinary skill in the field understood.
Expression or logic and/or step described otherwise above herein in flow charts, for example, being considered use
In the order list for the executable instruction for realizing logic function, may be embodied in any computer-readable medium, for
Instruction execution system, device or equipment (such as computer based system including the system of processor or other can be held from instruction
The system of row system, device or equipment instruction fetch and execute instruction) use, or combine these instruction execution systems, device or set
It is standby and use.For the purpose of this specification, " computer-readable medium " can any can be included, store, communicate, propagate or pass
Defeated program is for instruction execution system, device or equipment or the dress used with reference to these instruction execution systems, device or equipment
Put.The more specifically example (non-exhaustive list) of computer-readable medium includes following：Electricity with one or more wiring
Connecting portion (electronic device), portable computer diskette box (magnetic device), random access memory (RAM), read-only storage
(ROM), erasable edit read-only storage (EPROM or flash memory), fiber device, and portable optic disk is read-only deposits
Reservoir (CDROM).In addition, computer-readable medium can even is that the paper that can print described program on it or other are suitable
Medium, because can be for example by carrying out optical scanner to paper or other media, then into edlin, interpretation or if necessary with it
His suitable method is handled electronically to obtain described program, is then stored in computer storage.
It should be appreciated that each several part of the present invention can be realized with hardware, software, firmware or combinations thereof.Above-mentioned
In embodiment, software that multiple steps or method can be performed in memory and by suitable instruction execution system with storage
Or firmware is realized.Such as, if realized with hardware with another embodiment, following skill well known in the art can be used
Any one of art or their combination are realized：With the logic gates for realizing logic function to data-signal from
Logic circuit is dissipated, the application-specific integrated circuit with suitable combinational logic gate circuit, programmable gate array (PGA), scene can compile
Journey gate array (FPGA) etc..
Those skilled in the art are appreciated that to realize all or part of step that above-described embodiment method carries
Suddenly it is that relevant hardware can be instructed to complete by program, the program can be stored in a kind of computer-readable storage medium
In matter, the program upon execution, including one or a combination set of the step of embodiment of the method.
In addition, each functional unit in each embodiment of the present invention can be integrated in a processing module, can also
That unit is individually physically present, can also two or more units be integrated in a module.Above-mentioned integrated mould
Block can both be realized in the form of hardware, can also be realized in the form of software function module.The integrated module is such as
Fruit is realized in the form of software function module and as independent production marketing or in use, can also be stored in a computer
In read/write memory medium.
Storage medium mentioned above can be read-only storage, disk or CD etc..Although have been shown and retouch above
The embodiment of the present invention is stated, it is to be understood that above-described embodiment is exemplary, it is impossible to be interpreted as the limit to the present invention
System, those of ordinary skill in the art can be changed above-described embodiment, change, replace and become within the scope of the invention
Type.
Claims (10)
1. a kind of equipment records processing method, it is characterised in that for determining the classification belonging to each bar equipment record, with according to institute
The maintenance description in the unified equipment record of classification is stated, is comprised the following steps：
The first eigenvector in measurement value sensor generation primary vector space in being recorded according to each bar equipment, and according to each
Maintenance description in bar equipment record, generates the second feature vector in secondary vector space；
According to the mapping relations between the primary vector space and the secondary vector space, determine that each first eigenvector exists
The first map vector in secondary vector space, and each second feature vector in primary vector space second map to
Amount；
Choose second map vector to be added in the primary vector collection comprising the first eigenvector, and choose described
One map vector is added in the secondary vector collection comprising second feature vector；The second map vector chosen is corresponding
Second feature vector be belonging respectively to the similarity vectors of different vector spaces and cluster, the first map vector of selection is corresponding
The similarity vector that first eigenvector is belonging respectively to different vector spaces clusters；
Vector in the primary vector collection is clustered to obtain first object and is clustered, and in the secondary vector collection
Vector, which is clustered to obtain the second target, to cluster；
The similar cluster to cluster to the first object in clustering with second target is combined, and according to each equipment
Combination belonging to the first eigenvector of record, second feature vector, the first map vector and the second map vector, determines described
Classification belonging to equipment record.
2. equipment according to claim 1 records processing method, it is characterised in that in the record according to each bar equipment
First eigenvector in measurement value sensor generation primary vector space, and the maintenance in each bar equipment record describes,
The second feature vector in secondary vector space is generated, including：
A dimension using each measurement value sensor as vector, carries out vectorization, obtains the first eigenvector；Institute
Measurement value sensor is stated including at least one in temperature value, rotating speed and pressure value；
A dimension using each maintenance description as vector, carries out vectorization, obtains the second feature vector；The dimension
Guarantor's description includes at least one in maintenance time, device type, fault category, failure rank, vendor name and the source of failure.
3. equipment according to claim 2 records processing method, it is characterised in that described by each measurement value sensor
As a dimension of vector, vectorization is carried out, before obtaining the first eigenvector, is further included：
It is averaged to same sensor in measured value at different moments, obtains a kind of measurement value sensor.
4. equipment according to claim 1 records processing method, it is characterised in that described to choose second map vector
It is added in the primary vector collection comprising the first eigenvector, and chooses first map vector and be added to comprising described
In the secondary vector collection of second feature vector, including：
In primary vector space, the first eigenvector is clustered to obtain each primary vector clustered；In secondary vector
In space, the second feature vector is clustered to obtain each secondary vector clustered；
According to the mapping relations between primary vector space and secondary vector space, determine that primary vector clusters and gather with secondary vector
Similitude between cluster；
If the primary vector belonging to first eigenvector clusters, belonging to the first map vector corresponding with the first eigenvector
Secondary vector cluster between there are similitude, first map vector is added in the secondary vector collection；
If the secondary vector belonging to the second feature vector clusters, the second map vector corresponding with second feature vector
Affiliated primary vector cluster between there are similitude, second map vector is added in the primary vector collection.
5. equipment according to claim 4 records processing method, it is characterised in that described according to primary vector space and the
Mapping relations between two vector spaces, determine primary vector cluster and secondary vector cluster between similitude, including：
Obtain the mapping relations between primary vector space and secondary vector space；
According to the mapping relations, after the centroid vector that the primary vector clusters is mapped to the secondary vector space, meter
Calculate and the secondary vector space in the first distance for clustering between centroid vector of secondary vector；
According to the mapping relations, after the centroid vector that the secondary vector clusters is mapped to the primary vector space, meter
Calculate and the primary vector space in the second distance that clusters between centroid vector of primary vector；
According to first distance and the second distance, between determining that the primary vector clusters and the secondary vector clusters
Similitude.
6. equipment according to claim 5 records processing method, it is characterised in that the acquisition primary vector space and the
Mapping relations between two vector spaces, including：
By first eigenvector vsSubstitute into transfer function V'=σ (w*V+b) and be iterated computing, when taking different value to obtain w and b
The first map vector vd', and according to the first map vector vd' and corresponding second feature vector vdDetermine that loss function takes
Value；σ is S type functions；When loss function value is less than threshold value or while reaching maximum iteration stops iteration, during by stopping iteration
W and b values substitute into the transfer function, as the mapping relations that the primary vector spatial transformation is secondary vector space；
By second feature vector vdSubstitute into transfer function V'=σ (w*V+b) and be iterated computing, when taking different value to obtain w and b
The second map vector vs', and according to the second map vector vs' and corresponding first eigenvector vsDetermine that loss function takes
Value；σ is S type functions；When loss function value is less than threshold value or while reaching maximum iteration stops iteration, during by stopping iteration
W and b values substitute into the transfer function, as the mapping relations that the secondary vector spatial transformation is primary vector space.
7. equipment according to claim 1 records processing method, it is characterised in that it is described cluster to first object with it is described
Second target cluster in similar cluster be combined before, further include：
Obtain the mapping relations between primary vector space and secondary vector space；
According to the mapping relations, after the centroid vector that the first object clusters is mapped to the secondary vector space, meter
Calculate and the secondary vector space in the 3rd distance that clusters between centroid vector of the second target；
According to the mapping relations, after the centroid vector that second target clusters is mapped to the primary vector space, meter
Calculate and the primary vector space in the 4th distance that clusters between centroid vector of first object；
According to the 3rd distance and the 4th distance, between determining that the first object clusters and second target clusters
Similitude.
8. a kind of equipment record processing unit, it is characterised in that for determining the classification belonging to each bar equipment record, with according to institute
The maintenance description in the unified equipment record of classification is stated, including：
Generation module, for the fisrt feature in the measurement value sensor generation primary vector space in being recorded according to each bar equipment
Vector, and the maintenance in each bar equipment record describes, the second feature vector in generation secondary vector space；
Determining module, for according to the mapping relations between the primary vector space and the secondary vector space, determining each
First map vector of the first eigenvector in secondary vector space, and each second feature vector is in primary vector space
The second map vector；
Module is chosen, is added to the primary vector collection comprising the first eigenvector for choosing second map vector
It is interior, and choose first map vector and be added in the secondary vector collection comprising second feature vector；Second chosen
The similarity vector that the corresponding second feature vector of map vector is belonging respectively to different vector spaces clusters, and the first of selection reflects
The similarity vector that the corresponding first eigenvector of directive amount is belonging respectively to different vector spaces clusters；
Cluster module, clusters for being clustered to obtain first object to the vector in the primary vector collection, and to described
Vector in secondary vector collection, which is clustered to obtain the second target, to cluster；
Sort module, is combined for the similar cluster in clustering with second target that clusters to the first object, and
Belonging to first eigenvector, second feature vector, the first map vector and the second map vector recorded according to each equipment
Combination, determine the classification belonging to equipment record.
9. a kind of computer equipment, it is characterised in that including memory, processor and storage on a memory and can be in processor
The computer program of upper operation, when the processor performs the computer program, is realized such as any one of claim 1-7 institutes
The equipment record processing method stated.
10. a kind of non-transitorycomputer readable storage medium, is stored thereon with computer program, it is characterised in that the program
The equipment record processing method as any one of claim 1-7 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711397566.2A CN108009647B (en) | 2017-12-21 | 2017-12-21 | Device record processing method and device, computer device and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711397566.2A CN108009647B (en) | 2017-12-21 | 2017-12-21 | Device record processing method and device, computer device and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108009647A true CN108009647A (en) | 2018-05-08 |
| CN108009647B CN108009647B (en) | 2020-10-30 |
Family
ID=62060529
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711397566.2A Active CN108009647B (en) | 2017-12-21 | 2017-12-21 | Device record processing method and device, computer device and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108009647B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116935608A (en) * | 2023-09-18 | 2023-10-24 | 广州春光新能源科技发展有限公司 | Remote intelligent monitoring method and system for air energy equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106611052A (en) * | 2016-12-26 | 2017-05-03 | 东软集团股份有限公司 | Text label determination method and device |
| CN106776740A (en) * | 2016-11-17 | 2017-05-31 | 天津大学 | A kind of social networks Text Clustering Method based on convolutional neural networks |
| CN106776713A (en) * | 2016-11-03 | 2017-05-31 | 中山大学 | It is a kind of based on this clustering method of the Massive short documents of term vector semantic analysis |
| US20170212952A1 (en) * | 2013-04-29 | 2017-07-27 | Moogsoft Inc. | Chat driven interface for processing clusters of events from managed infrastructures |
-
2017
- 2017-12-21 CN CN201711397566.2A patent/CN108009647B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170212952A1 (en) * | 2013-04-29 | 2017-07-27 | Moogsoft Inc. | Chat driven interface for processing clusters of events from managed infrastructures |
| CN106776713A (en) * | 2016-11-03 | 2017-05-31 | 中山大学 | It is a kind of based on this clustering method of the Massive short documents of term vector semantic analysis |
| CN106776740A (en) * | 2016-11-17 | 2017-05-31 | 天津大学 | A kind of social networks Text Clustering Method based on convolutional neural networks |
| CN106611052A (en) * | 2016-12-26 | 2017-05-03 | 东软集团股份有限公司 | Text label determination method and device |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116935608A (en) * | 2023-09-18 | 2023-10-24 | 广州春光新能源科技发展有限公司 | Remote intelligent monitoring method and system for air energy equipment |
| CN116935608B (en) * | 2023-09-18 | 2023-12-05 | 广州春光新能源科技发展有限公司 | Remote intelligent monitoring method and system for air energy equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108009647B (en) | 2020-10-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101957760B1 (en) | System for estimating market price of real estate using sales cases determined based on similarity score and method thereof | |
| CN116431975B (en) | Environment monitoring method and system for data center | |
| US9852360B2 (en) | Data clustering apparatus and method | |
| CN112560355A (en) | Method and device for predicting Mach number of wind tunnel based on convolutional neural network | |
| CN105045863B (en) | A kind of method and system for Entities Matching | |
| LU500017B1 (en) | Fault diagnosis device for three-phase asynchronous motor based on digital twinning and transfer learning | |
| CN110147820A (en) | Recommended method, device, equipment and the storage medium of the additional oil mass of flight | |
| CN109100142B (en) | A kind of semi-supervised method for diagnosing faults of bearing based on graph theory | |
| KR20080084132A (en) | Method for extracting nonlinear time series prediction model using neural network with weighted fuzzy membership functions | |
| CN108090318A (en) | Sensing data generates and response processing storehouse | |
| CN111562541B (en) | Software platform for realizing electric energy meter detection data management by applying CART algorithm | |
| CN109271466A (en) | A kind of weather data analysis method based on hierarchical clustering Yu K mean algorithm | |
| CN106250918A (en) | A kind of mixed Gauss model matching process based on the soil-shifting distance improved | |
| CN108009647A (en) | Equipment record processing method, device, computer equipment and storage medium | |
| CN108280158B (en) | Non-rigid three-dimensional model retrieval method based on gradient direction accumulation thermonuclear characteristics | |
| CN117036781A (en) | Image classification method based on tree comprehensive diversity depth forests | |
| Li et al. | Real‐time approach for oscillatory stability assessment in large‐scale power systems based on MRMR classifier | |
| CN109886288A (en) | A kind of method for evaluating state and device for power transformer | |
| CN114490412A (en) | Three-dimensional CAD software performance measurement method and device based on self-subtraction reverse cloud generator | |
| CN109270010A (en) | A kind of plant regulation of absorbing dust capability detection method based on random forests algorithm | |
| CN114021619A (en) | Classification method and device based on graph structure evolution | |
| CN114757097A (en) | Line fault diagnosis method and device | |
| CN107710085A (en) | For the methods, devices and systems for the damaged condition for estimating electric installation | |
| Barzegkar-Ntovom et al. | Generic dynamic load modelling using cluster analysis | |
| Gao et al. | Fused machine learning based on Boosting algorithm for noise recognition of moving parts in sealed cavity |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |